Tuto part2

Sparse Coding and Dictionary Learning
for Image Analysis
Part II: Dictionary Learning for signal reconstruction

Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro

ICCV’09 tutorial, Kyoto, 28th September 2009

Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Dictionary Learning for signal reconstruction 1/43

What this part is about
The learning of compact representations of images
adapted to restoration tasks.
A fast online algorithm for learning dictionaries and
factorizing matrices in general.
Various formulations for image and video processing.


The Image Denoising Problem

y = xorig + w
measurements original image noise


Sparse representations for image restoration

y = xorig + w
measurements original image noise

Energy minimization problem - MAP estimation
E (x) = ||y − x||2
2 + Pr (x)
relation to measurements prior

Some classical priors
Smoothness λ||Lx||2
2
Total variation λ|| x||2
1
Wavelet sparsity λ||Wx||1
...


Sparsity and redundancy
Pr (x) = λ||α||0 for x ≈ Dα
 
    α[1]
α[2]
x =  d1 d2 · · · dp   .  . 
 . 
α[p]
x∈Rm D∈Rm×p
α∈Rp ,sparse


Designed dictionaries
[Haar, 1910], [Zweig, Morlet, Grossman ∼70s], [Meyer, Mallat,
Daubechies, Coifman, Donoho, Candes ∼80s-today]. . . (see [Mallat,
1999])
Wavelets, Curvelets, Wedgelets, Bandlets, . . . lets

Learned dictionaries of patches
[Olshausen and Field, 1997], [Engan et al., 1999], [Lewicki and
Sejnowski, 2000], [Aharon et al., 2006] , [Roth and Black, 2005], [Lee
et al., 2007]
1
min ||xi − Dαi ||2 + λψ(αi )
2
αi ,D∈C 2
i
sparsity
reconstruction
ψ(α) = ||α||0 (“ 0 pseudo-norm”)
ψ(α) = ||α||1 ( 1 norm)


Solving the denoising problem
[Elad and Aharon, 2006]
Extract all overlapping 8 × 8 patches yi .
Solve a matrix factorization problem:
n
1
min ||yi − Dαi ||2 + λψ(αi ),
2
αi ,D∈C 2
i=1 sparsity
reconstruction

with n > 100, 000
Average the reconstruction of each patch.


K-SVD: [Elad and Aharon, 2006]

Figure: Dictionary trained on a noisy version of the image
boat.


Inpainting, Demosaicking
1
min ||β ⊗ (yi − Dαi )||2 + λi ψ(αi )
D∈C,α 2 i 2
i

RAW Image Processing (see our poster)

White
balance.
Denoising
Black
substraction.
Conversion
to sRGB.
Demosaicking
Gamma
correction.


[Mairal, Bach, Ponce, Sapiro, and Zisserman, 2009c]


[Mairal, Sapiro, and Elad, 2008b]


Inpainting, [Mairal, Elad, and Sapiro, 2008a]


Sparse representations for video restoration
Key ideas for video processing
[Protter and Elad, 2009]
Using a 3D dictionary.
Processing of many frames at the same time.
Dictionary propagation.


Inpainting, [Mairal, Sapiro, and Elad, 2008b]

Figure: Inpainting results.


Color video denoising, [Mairal, Sapiro, and Elad, 2008b]

Figure: Denoising results. σ = 25


Optimization for Dictionary Learning

n
1
min ||xi − Dαi ||2 + λ||αi ||1
2
α∈R p×n 2
D∈C i=1

C = {D ∈ Rm×p s.t. ∀j = 1, . . . , p, ||dj ||2 ≤ 1}.

Classical optimization alternates between D and α.
Good results, but very slow!


[Mairal, Bach, Ponce, and Sapiro, 2009a]

Classical formulation of dictionary learning
n
1
min fn (D) = min l(xi , D),
D∈C D∈C n
i=1

where
1
l(x, D) = minp ||x − Dα||2 + λ||α||1 .
2
α∈R 2

Which formulation are we interested in?
n
1
min f (D) = Ex [l(x, D)] ≈ lim l(xi , D)
D∈C n→+∞ n
i=1

[Bottou and Bousquet, 2008]: Online learning can
handle potentially inﬁnite or dynamic datasets,
be dramatically faster than batch algorithms.

Require: D0 ∈ Rm×p (initial dictionary); λ ∈ R
1: A0 = 0, B0 = 0.
2: for t=1,. . . ,T do
3: Draw xt
4: Sparse Coding

1
αt ← arg min ||xt − Dt−1 α||2 + λ||α||1 ,
2
α∈Rp 2

5: Aggregate suﬃcient statistics
At ← At−1 + αt αT , Bt ← Bt−1 + xt αT
t t
6: Dictionary Update (block-coordinate descent)
t
1 1
Dt ← arg min ||xi − Dαi ||2 + λ||αi ||1 .
2
D∈C t 2
i=1

7: end for

Which guarantees do we have?
Under a few reasonable assumptions,
ˆ
we build a surrogate function ft of the expected cost f verifying
ˆ
lim ft (Dt ) − f (Dt ) = 0,
t→+∞

Dt is asymptotically close to a stationary point.

Extensions (all implemented in SPAMS)
non-negative matrix decompositions.
sparse PCA (sparse dictionaries).
fused-lasso regularizations (piecewise constant dictionaries)


Experimental results, batch vs online

m = 8 × 8, p = 256

Experimental results, batch vs online

m = 12 × 12 × 3, p = 512

Inpainting a 12-Mpixel photograph


Extension to NMF and sparse PCA
[Mairal, Bach, Ponce, and Sapiro, 2009b]

NMF extension
n
1
min ||xi − Dαi ||2 s.t. αi ≥ 0,
2 D ≥ 0.
α∈Rp×n 2
D∈C i=1

SPCA extension
n
1
min ||xi − Dαi ||2 + λ||α1 ||1
2
α∈Rp×n 2
D∈C i=1

C = {D ∈ Rm×p s.t. ∀j ||dj ||2 + γ||dj ||1 ≤ 1}.
2


Faces: Extended Yale Database B

(a) PCA (b) NNMF (c) DL


Faces: Extended Yale Database B

(d) SPCA, τ = 70% (e) SPCA, τ = 30% (f) SPCA, τ = 10%


Natural Patches

(a) PCA (b) NNMF (c) DL


Natural Patches

(d) SPCA, τ = 70% (e) SPCA, τ = 30% (f) SPCA, τ = 10%


Summary of this part
The dictionary learning framework leads to
state-of-the-art results for many image . . .
. . . and video processing tasks.
Online learning techniques are well-suited for this
problem and allows training sets with millions of
patches.


References I
M. Aharon, M. Elad, and A. M. Bruckstein. The K-SVD: An algorithm for designing
of overcomplete dictionaries for sparse representations. IEEE Transactions on
Signal Processing, 54(11):4311–4322, November 2006.
L. Bottou and O. Bousquet. The trade-offs of large scale learning. In J.C. Platt,
D. Koller, Y. Singer, and S. Roweis, editors, Advances in Neural Information
Processing Systems, volume 20, pages 161–168. MIT Press, Cambridge, MA, 2008.
M. Elad and M. Aharon. Image denoising via sparse and redundant representations
over learned dictionaries. IEEE Transactions on Image Processing, 54(12):
3736–3745, December 2006.
K. Engan, S. O. Aase, and J. H. Husoy. Frame based signal compression using
method of optimal directions (MOD). In Proceedings of the 1999 IEEE
International Symposium on Circuits Systems, volume 4, 1999.
A. Haar. Zur theorie der orthogonalen funktionensysteme. Mathematische Annalen,
69:331–371, 1910.
H. Lee, A. Battle, R. Raina, and A. Y. Ng. Efficient sparse coding algorithms. In
B. Sch¨lkopf, J. Platt, and T. Hoffman, editors, Advances in Neural Information
o
Processing Systems, volume 19, pages 801–808. MIT Press, Cambridge, MA, 2007.


References II
M. S. Lewicki and T. J. Sejnowski. Learning overcomplete representations. Neural
Computation, 12(2):337–365, 2000.
J. Mairal, M. Elad, and G. Sapiro. Sparse representation for color image restoration.
IEEE Transactions on Image Processing, 17(1):53–69, January 2008a.
J. Mairal, G. Sapiro, and M. Elad. Learning multiscale sparse representations for
image and video restoration. SIAM Multiscale Modelling and Simulation, 7(1):
214–241, April 2008b.
J. Mairal, F. Bach, J. Ponce, and G. Sapiro. Online dictionary learning for sparse
coding. In Proceedings of the International Conference on Machine Learning
(ICML), 2009a.
J. Mairal, F. Bach, J. Ponce, and G. Sapiro. Online learning for matrix factorization
and sparse coding. ArXiv:0908.0050v1, 2009b. submitted.
J. Mairal, F. Bach, J. Ponce, G. Sapiro, and A. Zisserman. Non-local sparse models
for image restoration. In Proceedings of the IEEE International Conference on
Computer Vision (ICCV), 2009c.
S. Mallat. A Wavelet Tour of Signal Processing, Second Edition. Academic Press,
New York, September 1999.


References III
B. A. Olshausen and D. J. Field. Sparse coding with an overcomplete basis set: A
strategy employed by V1? Vision Research, 37:3311–3325, 1997.
M. Protter and M. Elad. Image sequence denoising via sparse and redundant
representations. IEEE Transactions on Image Processing, 18(1):27–36, 2009.
S. Roth and M. J. Black. Fields of experts: A framework for learning image priors. In
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition
(CVPR), 2005.


Tuto part2

Recommandé

Recommandé

Contenu connexe

Tendances

Tendances (18)

Similaire à Tuto part2

Similaire à Tuto part2 (19)

Dernier

Dernier (20)

Tuto part2